Professional Documents
Culture Documents
Netherlands
(2)
(3)
(3)
ABSTRACT
This study addresses risk assessment for sediment disasters based on watershed-wide estimations of soil moisture
resulting from long-term hydrologic processes. Sediment disasters have recently occurred throughout Japan due to
heavy rainfall and rapid snowmelt. Climate change is expected to exacerbate such disasters in snowy regions, due to
global warming.
Quantitative evaluation for risk of landslides such as those that have occurred at Nakayama Pass in Sapporo, Northern
Japan, during the snowmelt season has remained an issue. A method for determining the soil moisture at potential
sediment disaster sites was proposed and applied to the Nakayama Pass disaster. This site is in the Hoheikyo Dam
watershed and thus is influenced by its hydrologic characteristics.
We propose methods to quantitatively estimate soil moisture, which is an important factor in sediment disasters, by
using the Soil Water Index (SWI) and Water Storage (WS) estimated as THE water level of the tank model when rainfall
and snowmelt are given. Model parameters of the SWI are uniform in all Japan. In contrast, parameters of WS can be
changed in depending on watershed characteristics. WS in each 1-km by 1-km mesh is estimated as the water level of
the tank model when rainfall and snowmelt are given. Then, the total outflow from the watershed is estimated by
synthesizing the outflow for each mesh using a channel-routing method based on kinematic waves. The validity of the
estimated WS is indirectly confirmed by the reproducibility of total outflow from the dam catchment that includes the
disaster point.
Amounts of SWI and WS at the disaster site were estimated, and then it was found that the landslides of 2012 and 2000
had occurred under the condition of maximum SWI or maximum WS resulting from heavy rainfall, combined with
snowmelt. SWI had slightly underestimated the risk. Thus, this study suggested that the amount of WS based on
hydrologic cycle in a catchment area influences large-scale slope disasters such as landslides. It is believed that the risk
of slope disasters resulting from high soil moisture content can be predicted by incorporating appropriate weather data
into the distributed hydrologic model, which is a model that can consider rainfall as well as snowmelt.
Keywords: Sediment disaster, Snowmelt, Hydrologic processes, Soil Water Index, Water Storage
1.
INTRODUCTION
Slope disasters have recently occurred throughout Japan due to heavy rainfall and rapid snowmelt. Climate change is
expected to exacerbate such disasters in snowy regions, due to global warming. The quantitative evaluation for risk of
landslides such as those that have occurred at Nakayama Pass in Sapporo, Northern Japan, during the snowmelt
season has remained an issue. A method for determining the soil moisture at potential slope disaster sites was
proposed and applied to the Nakayama Pass disaster that resulted from rainfall during the snowmelt season. The
investigation site is on Route 230 connecting Sapporo, the capital of Hokkaido Prefecture, to Southern Hokkaido. This
site is in the Hoheikyo Dam watershed and thus is influenced by its hydrologic characteristics.
Some reports have proposed methods for evaluating the risk of landslide disasters in Japan (Osanai et al. (2010), MLIT,
JMA and NILIM (2005)). Toward establishing a method for using meteorological factors in estimating the location and
timing of landslides, many researchers have proposed various methods, because many different factors influence
landslides. Intense rainfall is known as the primary meteorological factor in triggering landslides. Antecedent
precipitation is recognized as a secondary landslide triggering meteorological factor. Wieczorek and Glade (2005)
summarized the criteria for debris-flow triggering threshold used in various countries. Their summary shows that various
rainfall indices have been used for determining the criteria. For example, Glade et al. (2000) shows that rainfall
probability thresholds can be established by applying an Antecedent Daily Rainfall model that uses two variables:
antecedent daily rainfall index and daily rainfall. A method similar in concept to that of Glade et al. (2000) has been used
in Japanese early-warning systems that determine the criterion of landslides occurrence using two rainfall indices.
1
In the place of antecedent rainfall as a secondary trigger of sediment disaster, we have proposed methods based on
effective rainfall and those using the soil water index (SWI), which is obtained by measuring soil moisture and using a
tank model. SWI is the method that is used in the issuance of "landslide disaster warning information" (JMA (2012)) by
the Japan Meteorological Agency and prefectural governments. SWI is pertinent as disaster information regarding
landslides that are caused by heavy rainfall only. However the information provision system does not address disasters
whose causes include snowmelt. Thus, the establishment of methods for evaluating landslide risk that take snowmelt
into consideration has been a research task.
To evaluate landslide occurrence risk that includes both rainfall and snowmelt, we proposed the method that uses 60minute rainfall-snowmelt and SWI. Moreover, an attempt was made to quantitatively estimate soil moisture, which is an
important factor in slope disasters, by using the water storage (WS) for each mesh as estimated in the distributed
hydrologic model of the watershed. In this hydrologic model, snowmelt is estimated on the basis of the heat balance
between the snowpack layer and the atmosphere. WS in each 1-km by 1-km mesh is estimated as the water level of the
tank model when rainfall and snowmelt are provided. Then, the total outflow from the watershed is estimated by
synthesizing the outflow for each mesh using a channel-routing method based on kinematic waves. The validity of the
estimated WS is indirectly confirmed by the reproducibility of total outflow from the dam catchment that includes the
disaster point.
Amounts of SWI and WS at the disaster site were estimated, and then it was found that the landslides of 2012 and 2000
had occurred under the condition of maximum SWI or maximum WS resulting from heavy rainfall, combined with
snowmelt. SWI had slightly underestimated the risk. Thus, this study suggested that the amount of WS based on
hydrologic cycle in a catchment area influences large-scale slope disasters such as landslides. It is believed that the risk
of slope disasters resulting from high soil moisture content can be predicted by incorporating appropriate weather data
into the distributed hydrologic model, which is a model that can consider rainfall as well as snowmelt, so that it will be of
practical use to disaster mitigation, such as the closure of roads.
2.
METHODS
2.1
The locations of disaster sites are shown in Figure 1. Landslides due to snowmelt as well as rainfall in 2000 and 2012 at
Nakayama Pass on National Road 230, which connects Sapporo and Southern Hokkaido, were specifically severe as
shown in Photo 1. The road was completely closed for the first time since its opening in 1969. On May, 2012, the
landslides occurred along the range of the road at about 10m in width and about 86m in length at KP40.6. At KP40.8, a
slope failure occurred and about 13,000 m 3 of soil collapsed in the range of about 40m in road length and about 110m in
transverse width. It took 20 days after the occurrence of the disaster for just one lane of the section to reopen.
Sightseeing activities for people during the "Golden Week" holiday period in Japan, as well as the logistics, were affected
by the long period of road closure.
Figure 2 shows changes in the annual greatest daily rainfall from 1975 to 2013 measured by the management office of
Hoheikyo Dam. The bar indicates the annual greatest daily rainfall, and the circle indicates the greatest daily rainfall
during the period from April to May, which is the snowmelt season in the subject area. It can be seen that the greatest of
the yearly greatest daily rainfalls occurred in 1981. Heavy rainfall events that surpassed the daily rainfall of the May 4th,
2012 event have occurred many times in the past. When the data only from the snowmelt season are analyzed, the daily
rainfall of the Nakayama Pass disaster is seen to be not the greatest in the history of that pass. The scale of rainfall
does not show any specificity to the disaster in question.
200
Yearly largest
Largest in April and May
150
100
50
0
1975
1980
1985
1990
1995
Year
2000
2005
2010
2015
Figure 2. Changes in the largest daily rainfall (measured at the management office of Hoheikyo Dam)
700
Yearly largest inflow
600
Inflow (m3/s)
500
May 12th, 2000
400
300
200
100
0
1975
1980
1985
1990
1995
Year
2000
2005
2010
2015
Figure 3. Yearly greatest inflow and the greatest inflow during April and May for Hoheikyo Dam
Thus it is necessary to consider snowmelt as a part of water that is supplied to the soil during April and May. Snowmelt
water and rain water infiltrate the soil and eventually flow out to rivers. The amount of water that is supplied to the soil
corresponds to the amount of water discharged by the river. Therefore, the runoff conditions at Hoheikyo Dam, which
receives the water of the watershed that included the site of disaster, were investigated (Figure 3). The greatest inflow for
each year, and that for April and May of each year, were extracted from the hourly inflow data and plotted. The bar in the
figure indicates the greatest value for the year, and the circle indicates the greatest value for April and May. The greatest
inflow was observed in 1981, which is the year when the greatest daily rainfall was observed. The next-greatest inflow
was observed in 2011. The discharges in 2000 and 2012 were very great, but neither was the largest flood discharge on
record. The water supplied to the soil at the time of the subject disaster is thought to be of a scale which had occurred in
the past. Based on this analysis, it is thought that snowmelt water was a major contributor to the disasters of May 12th,
2000 and May 4th, 2012.
2.2
Hourly snowmelt was estimated by a method based on the heat balance between atmosphere and snow surface
proposed by Kondo and Yamazaki (1990) (1992). The basic equations are as follows:
cs s
[ Z (T0 Ts ) Z n (T0 Tsn )] W0 s l f ( Z Z n ) M 0 t Gt
2
[1]
G (1 ) S ( L T 4 ) H lE
[2]
( L Tsn4 ) H lE s
T0 Tsn
0
Zn
[3]
where, equation (1) gives the energy balance under the assumption that during t, the freezing depth changed from Z(m)
to Zn(m) and that the snow surface temperature changed from Ts(K) to Tsn(K). Equation (2) calculates the heat flux from
the atmosphere. Equation (3) gives the heat balance in an area of infinitesimally thin snowpack layer. Here, cs is the
specific heat of snowpack (J/kg/K), s is the density of snowpack (kg/m 3), T0 is 0C (273K), lf is the latent heat of melting
(J/kg), W0 is the moisture percentage of snowpack, M0 is snowmelt energy (W/m 2), G is the energy provided from the
atmosphere to the snow surface (W/m 2), is the albedo of the snow surface, S is the flux of solar radiation (W/m 2), L is
the flux of downward longwave radiation, is the Stefan-Boltzmann constant (W/m 2/K4), T is air temperature (K), H is
the flux of sensible heat (W/m 2), lE is the flux of latent heat (W/m 2), is the emissivity, s is the thermal conductivity of
snowpack (W/m/K). Of these, the snowpack density, the albedo, and the flux of downward longwave radiation are
parameterized by the air temperature and the relative humidity (Nakatsugawa et al. (1995)). The snowmelt energy (M0)
calculated from the above simultaneous equations is divided by the latent heat of melting and by the water density to
calculate the water equivalent of snowmelt (mm).
Hourly rainfall was also provided by the Automated Meteorological Data Acquisition System (AMeDAS) station in the
vicinity of the disaster site. The above snowmelt and rainfall were input to a three-cascade tank model (Figure 4) that is
adopted by the Japan Meteorological Agency (JMA), and then SWI was estimated. Here we applied the parameters of
the tank model that are currently employed in JMA.
Snowmelt
Snowpack
1km1km
Rainfall + Snowmelt
- Evapotranspiration
Tank model
Figure 5. Schematic of Water Storage (WS) based on the hydrologic processes in an entire watershed
2.2.2 Water storage (WS) based on the hydrologic processes in the entire watershed
Also, an attempt was made to consider the water stored in the soil resulting from the hydrologic processes in the entire
watershed (hereinafter: water storage (WS)) as an index of assessment. Then, the conditions of the disaster in question
and those of flood in the past that occurred at the time of heavy rainfall events are compared by focusing on amount of
water storage.
The hydrological variables (rainfall, snowmelt, evapotranspiration and runoff) which constitute the water cycle of a
watershed are estimated as a whole and are examined for accuracy of the estimated WS based on river discharge. In
estimating WS, it is important to reproduce the long-term water budget, including the processes of snowpack and
snowmelt. Calculation using mesh units is necessary, because the hydrological variables have a spatial distribution.
Thus, we applied the watershed hydrologic process model (Long-term Hydrologic Assessment model considering Snow
process (LoHAS)) in order to estimate WS as schematically shown in Figure 5. This method is based on the two-layer
model (Figure 6) proposed by Yamazaki et al. (1991) and it is modified by Nakatsugawa et al. (2004) and Usutani et al.
(2006) as follows:
fv R 1 f v Tv 4 QM QB QR Tg 4 H g lE g
[4]
1 f v R Tg 4 2 1 fv Tv 4 H v l Ev I
[5]
where, equations [4] and [5] are heat balance equations for the ground surface (i.e., the atmospheric layer from the
ground surface to the lowest limit of the canopy) and vegetation cover (i.e., the canopy layer), respectively. fv is
transmissivity to radiation of the vegetation cover, R is downward net radiation (W/m 2), QM is heat flux expended to
melt snow (W/m 2), QB is heat flux supplied to the ground (W/m 2), QR is heat flux supplied to the ground surface by
rainfall (W/m 2), Hg and Hv are sensible heat fluxes from the ground surface and vegetation cover, respectively (W/m 2),
lEg and lEv are latent heat fluxes from the ground surface and vegetation cover, respectively (W/m 2), lI is latent heat flux
associated with evaporation intercepted by the vegetation cover (W/m 2), Tg and Tv are representative temperatures at the
ground surface and vegetation cover, respectively (K), is emissivity (1.00 at ground surface, 0.97 at the ground snow
surface), and is the Stefan-Boltzmann constant (5.6710-8 W/m 2/K4). From these equations, representative
temperatures (Tg and Tv) and the heat flux expended to melt snow QM are obtained. The sum of QM on the snow surface
and QB at the bottom of the snowpack is divided by the latent heat of melting and by the water density to calculate the
water equivalent of snowmelt (mm). Sensible and latent heat fluxes are calculated by means of bulk equations (6) and
(7).
H g C p C HgU Tg T , H v C p CHvU Tv T
0.622
p
[6]
0.622
p
[7]
where, Cp is the specific heat at constant pressure (1,004J/K/Kg), is the density of air (kg/m ), U is the wind speed
(m/s) at a specific height above ground, T is the air temperature (K) at a specific height above ground, CHg and CHv are
bulk-transfer coefficients at the ground surface and in the vegetation layer, respectively, g and v are evaporation
effectiveness factors at the ground surface and in the vegetation layer, respectively, esat is saturated water vapor
pressure (hPa), e is water vapor pressure (hPa) at the specific elevation above ground, and p is atmospheric pressure on
the ground (hPa). Then, the values for bulk transfer coefficient and evaporation efficiency are decided for each mesh
according to the ground cover per 1-km mesh. The bulk transfer coefficient and evaporation efficiency for each ground
cover are chosen by referring to values in the literature (Kondo (1994)), and adjustments for both values are repeated to
achieve the best agreement with the water budget of the entire basin. The sum of lEg, lEv and lI is divided by the latent
heat of evaporation to calculate the amount of total evapotranspiration (mm). Thus, daily-unit hydrological variables
(rainfall, snowfall, snow cover, snowmelt and evapotranspiration) on each mesh of 1 km by 1 km can be estimated by
using this model.
Inflow to the dam reservoir was estimated by combining a river runoff routing model that is based on the kinematic wave
method with a slope runoff model for the relevant mesh that is based on the tank model. Calculation of runoff is
performed for each mesh in accordance with the input values for the amounts of rainfall, snowmelt and
evapotranspiration. For each mesh, the discharge from the upstream mesh is calculated by channel routing (channelrouting model), as well as the runoff that occurs within the mesh (slope runoff model). The rates of flow at the tail end of
the mesh are added, and that sum is the runoff into the downstream mesh. Channel routing is calculated from the
following equation, which is a modified kinematic wave equation.
Q 5 i 0.3Q 0.4 Q
0
t 3 n 0.6 B 0.4 x
[8]
where, Q is the discharge (m /s), i is the gradient, n is the roughness coefficient and B is the river width (m). The
gradient i is determined by extracting the lowest elevation in the mesh to be calculated and that of the mesh downstream,
and then obtaining the difference between the two elevations. To obtain the slope runoff within a mesh, the tank model
5
shown in Figure 5 is used. The parameter values in the tank model probably depend on the characteristics of the
watershed, e.g., the geological condition.
Incoming
Net Radiation
R
Outgoing
Net Radiation
(1-fv) Tv
Sensible
Heat
Hg Hv
T Atmosphere
Latent
Heat
Heat Flux
lEg lEv lI Due to rain
Tv Vegetation
Layer
(1-fv)Tg4
(1-fv)R
fvR
fvTg4
(1-fv)Tv4
QG=QM+QB
QR
Tg4
Heat Flux QM
for Snowmelt
Heat Flux QB
for Soil
Ts Snowpack
Tg Soil
SWI 10/05/0014/05/00
SWI 02/05/1206/05/12
WS 10/05/0014/05/00
WS 02/05/1206/05/12
50
40
30
20
10
0
0
50
100
150
200
250
Water amount (SWI and WS) mm)
300
Figure 7. Risk evaluation for sediment disaster occurrence based on the snake line (SL)
(Disaster occurrence risk is determined to be high when the SL exceeds the Critical Line (CL).)
300
250
200
150
100
50
0
0
100
200
300
400
500
600
Figure 9. The rainfall and snowmelt of the week before the day with the yearly largest water storage (WS), and those of the day with the yearly
largest WS
Figure 10 compares the storages of 1981 (the year with the greatest daily rainfall), 2000 (the large-scale snowmelt flood)
and 2012. The figures indicate that WS on the day before the disaster of May 4th, 2012 increased to about 250mm
because snowmelt had continued for many days. Therefore, it is thought that WS had reached the highest level, even
though the rainfall and snowmelt on the relevant day only were not the greatest values ever recorded. The high state of
WS tends to continue longer in snowmelt cases, such as those of 2000 and 2012, than in only heavy rain cases, such as
7
that of 1981. Moreover WS in the lower tank relating to subsurface and groundwater seems comparably high in the
snowmelt cases.
The results of disaster risk assessment using WS are shown as solid lines with square and circle in Figure 7. Snake
lines of 2000 and 2012 both exceed CL. Hence, WS is useful for evaluating disaster risk, because it is based on
hydrologic processes that depend on the watershed characteristics.
Since SWI and WS conceptually indicate the soil moisture, the physical meaning is not clear. Hence there is the need to
properly identify the parameters corresponding to the geological characteristics, such as thickness and permeability of
the soil layer. If there were archives of the hydrological data and geologic data that are required for estimating SWI and
WS, then the risk of sediment disasters such as landslides could be determined at any point.
3.
CONCLUSIONS
2012